AbsaOSS / spot

Aggregate and analyze Spark history, export to elasticsearch, visualize and monitor with Kibana.
Apache License 2.0
5 stars 0 forks source link

Add summary metrics including driver resources #34

Closed DzMakatun closed 3 years ago

DzMakatun commented 3 years ago

Add summary metrics which include resources allocated to driver. It is needed for proper monitoring of cluster load and total resource usage by application.

suggested metrics:

DzMakatun commented 3 years ago

when not explicitly set, driver memory and cores are missing in sparkProperties

attempt.aggs.allexecutors.driver.totalCores is always 0 when master yarn. For master local it shows all cores at the edge node.

so far it looks like there is no reliable way to get this data from Spark History

DzMakatun commented 3 years ago

attempt.environment.sparkProperties.spark.executor.memory is also missing when not set explicitly in Spark submit.

DzMakatun commented 3 years ago

To properly estimate total resource allocation, including driver and executor memory integration with external service e.g. YARN is needed.